The present status of the chemotherapy of brain tumors is disappointingly limited. Delivery of tumor toxic agents into the brain in sufficient concentration and with acceptably low side effects is severely limited by the limited diffusion of these agents into brain and brain tumors and by their systemic toxicity. This project proposes to study a new local delivery system employing semipermeable membranes which could be implanted into and around the tumor (usually at the time of primary tumor surgery). Membranes are now available which are non-toxic to brain and impermeable to bacteria, and which allow slow and constant diffusion of a variety of chemical agents directly into the region of the tumor. A series of chemotherapeutic drugs have been screened and screening will continue with other agents. A series of four steps in testing will be utilized. First, the rate of diffusion of each agent through the membrane will be quantitated in vitro in a water bath. Second, agents which diffuse through the membrane will be tested for short-term effect against a tissue-culture preparation of human glioblastoma. Third, each agent will be tested for long-term effect by diffusion through chronically implanted membranes against subcutaneous mouse ependymoma implants. And finally, each agent will be tested histologically and by radioautography in the brains of dogs to determine toxicity and depth of penetration into normal brain. Any agent which can be shown to be safe and effective when diffused in this fashion can then be considered for possible future clinical trials in human brain tumor patients.